preprocessing_ROI.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "8e146f34-847a-4370-93dd-6e6efbaad913",
   "metadata": {},
   "source": [
    "# Preprocessing code"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "456bd615-b312-41a7-84f5-5db03801a938",
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import glob\n",
    "from skimage import transform, util\n",
    "import nibabel as nib\n",
    "import tqdm\n",
    "from PIL import Image\n",
    "from skimage.restoration import denoise_tv_chambolle\n",
    "from skimage.measure import regionprops, label\n",
    "import matplotlib as mpl\n",
    "%matplotlib inline\n",
    "\n",
    "def get_ROI(im,mask,res):\n",
    "    # Give as an input an image (in the form of a dictionary with labels 'img'and 'mask) \n",
    "    # and outputs an ROI centred around the centre of mass of the heart. \n",
    "    # Second input is the resolutoin (mm/pixel) of the input image.\n",
    "    \n",
    "    # # split input into image and mask\n",
    "    # im = image['img']                          \n",
    "    # mask = image['mask']\n",
    "    \n",
    "    # find target size of ROI, based on pixel size (mm/pixel) and maximum expected size of heart (doi: 10.15171/jcvtr.2016.25)\n",
    "    # nr = np.round(16/res[0]*10/2).astype(int) # number of rows/2\n",
    "    # nc = np.round(16/res[1]*10/2).astype(int) # number of columns/2\n",
    "    nr = int(128/2) # target size is 128 by 128 pixels\n",
    "    nc = int(128/2)\n",
    "    mask_single_class = np.copy(mask)\n",
    "    # mask_single_class = np.expand_dims(mask_single_class, axis=0)\n",
    "    # print(mask_single_class)\n",
    "    # print(np.shape(mask_single_class))\n",
    "    # print(np.where(mask_single_class!=0))\n",
    "    # print(np.unique(mask_single_class))\n",
    "    mask_single_class[mask_single_class!=0] = 1\n",
    "    if len(np.unique(mask_single_class))==1:                                            # if there is no heart in the image, just use the central part of the image as an ROI\n",
    "        labels = label(mask_single_class,background=1)                                  # find all connected components\n",
    "        largestCC = labels == np.argmax(np.bincount(labels.flat)[1:])+1    # largest connected components\n",
    "        mask = largestCC*1                                                 # True --> 1, False --> 0     \n",
    "        properties = regionprops(mask)                                     # Geometrical properties of image\n",
    "        com = np.round(properties[0].centroid).astype(int)                 # Centre of mass = centre of image in this specific case\n",
    "        com_r = com[0]                                                     # row index\n",
    "        com_c = com[1]                                                     # column index\n",
    "        ROI_mask = mask[(com_r-nr):(com_r+nr),(com_c-nc):(com_c+nc)]     # ROI mask --> traget size 128*128 pixels\n",
    "        ROI_mask = 0*ROI_mask\n",
    "        ROI = im[(com_r-nr):(com_r+nr),(com_c-nc):(com_c+nc)]            # ROI of grey scale image\n",
    "\n",
    "    else:                                                                  # if heart is present\n",
    "        # Find Largest connected component:\n",
    "        labels = label(mask_single_class)                                  # find all connected components\n",
    "        largestCC = labels == np.argmax(np.bincount(labels.flat)[1:])+1    # largest connected components\n",
    "        largestCC = largestCC*1                                            # True --> 1, False --> 0\n",
    "        properties = regionprops(largestCC)                                # Geometrical properties of largest connected component\n",
    "        com = np.round(properties[0].centroid).astype(int)                 # Centre of mass\n",
    "        com_r = com[0]                                                     # row index\n",
    "        com_c = com[1]                                                     # column index\n",
    "        ROI_mask = mask[(com_r-nr):(com_r+nr),(com_c-nc):(com_c+nc)]\n",
    "        # print(np.unique(ROI_mask))\n",
    "        ROI = im[(com_r-nr):(com_r+nr),(com_c-nc):(com_c+nc)]    \n",
    "    # plt.figure()\n",
    "    # plt.imshow(np.squeeze(ROI),'gray')\n",
    "    # overlay_mask = np.ma.masked_where(ROI_mask==0, ROI_mask)\n",
    "    # plt.imshow(overlay_mask, 'Reds', alpha = 0.5, clim=[0,1], interpolation='nearest')\n",
    "    # plt.scatter(x=nc, y=nr, c='g', s=40)\n",
    "    # plt.figure()\n",
    "    \n",
    "    return ROI,ROI_mask\n",
    "\n",
    "\n",
    "def normalize_img(img):\n",
    "    # Warning for when dividing NaN value??\n",
    "    norm_img = np.divide(img,np.max(img))\n",
    "    return norm_img\n",
    "\n",
    "\n",
    "def crop_pad_resize(image, nx, ny):\n",
    "    x, y = image.shape\n",
    "\n",
    "    # difference in nr of pixels (divide by 2 since we have 2 sides)\n",
    "    x_s = (x - nx) // 2\n",
    "    y_s = (y - ny) // 2\n",
    "    x_c = (nx - x) // 2\n",
    "    y_c = (ny - y) // 2\n",
    "\n",
    "    if x > nx and y > ny:\n",
    "        # if image is larger in both dimensions cut a slice\n",
    "        slice_cropped = image[x_s:x_s + nx, y_s:y_s + ny]\n",
    "\n",
    "    else:\n",
    "        # if one dim is smaller fill that side up with 0's\n",
    "        slice_cropped = np.zeros((nx, ny))\n",
    "\n",
    "        if x <= nx and y > ny:\n",
    "            # fill up x direction with 0's, cut in x direction\n",
    "            slice_cropped[x_c:x_c + x, :] = image[:, y_s:y_s + ny]\n",
    "        elif x > nx and y <= ny:\n",
    "            # fill up y direction with 0's, cut in y direction\n",
    "            slice_cropped[:, y_c:y_c + y] = image[x_s:x_s + nx, :]\n",
    "        else:\n",
    "            # if dimensions are as desired, keep the original slice\n",
    "            slice_cropped[x_c:x_c + x, y_c:y_c + y] = image[:, :]\n",
    "\n",
    "    return slice_cropped\n",
    "\n",
    "\n",
    "def preprocess(input_folder, target_resolution, target_size, denoise=False, alphaTV=0.1):\n",
    "    '''\n",
    "    This function preprocesses ACDC data. It crops all images to the same size,\n",
    "    transforms everything to the same resolution and normalizes the images.\n",
    "    It automatically makes the folder where preprocessed data is written to,\n",
    "    in the same format as the ACDC data is given. The images are in PNG-format.\n",
    "    \n",
    "    input_folder: the folder where raw ACDC data is located.\n",
    "    target_resolution: desired resolution, should be a tuple with 2 items (x- and y-dimensions).\n",
    "    target_size: desired size. Should be a tuple wiht 2 items (x- and y-dimensions).\n",
    "    '''\n",
    "    corrupted_files = []\n",
    "    nx, ny = target_size\n",
    "    data_folder = input_folder\n",
    "    \n",
    "    if denoise:\n",
    "        foldername = 'preprocessed_ROI_TV005_v3'\n",
    "    else:\n",
    "        foldername = 'preprocessed_ROI_v3'\n",
    "    \n",
    "    if not os.path.exists(foldername):\n",
    "        os.mkdir(foldername)\n",
    "    else:\n",
    "        print(foldername+' folder already exists. Continuing regardless.')\n",
    "    \n",
    "    # Loop over train and test folders\n",
    "    for train_test in ['training', 'testing']:\n",
    "\n",
    "        input_folder = os.path.join(data_folder, train_test)\n",
    "        len_inp = len(input_folder)+1\n",
    "        \n",
    "        # Make train and test folders in preprocessed folder\n",
    "        if not os.path.exists(os.path.join(foldername+'/', train_test)):\n",
    "            os.mkdir(os.path.join(foldername+'/', train_test))\n",
    "        else:\n",
    "            print('T'+train_test[1:]+' folder already exists. Continuing regardless.')\n",
    "        \n",
    "        # Loop over patient folders\n",
    "        for folder in os.listdir(input_folder):\n",
    "            \n",
    "            if folder != '.ipynb_checkpoints':  # Sometimes trouble with automatically made files\n",
    "\n",
    "                folder_path = os.path.join(input_folder, folder)\n",
    "                \n",
    "                # Make patient folders in preprocessed folder\n",
    "                if not os.path.exists(os.path.join(foldername+'/'+train_test, folder_path[len_inp:])):\n",
    "                    os.mkdir(os.path.join(foldername+'/'+train_test, folder_path[len_inp:]))\n",
    "                else:\n",
    "                    print('Folder for '+folder_path[len_inp:]+' already exists. Continuing regardless.')\n",
    "                \n",
    "                if os.path.exists(foldername+'/'+train_test+'/'+folder_path[len_inp:]+'/.ipynb_checkpoints'):\n",
    "                    os.rmdir(foldername+'/'+train_test+'/'+folder_path[len_inp:]+'/.ipynb_checkpoints')\n",
    "                    \n",
    "                lst = os.listdir(foldername+'/'+train_test+'/'+folder_path[len_inp:])\n",
    "                \n",
    "                if len(lst) == 0:  # Only create files if the designated folder is empty\n",
    "                    \n",
    "                    for file in glob.glob(os.path.join(folder_path, 'patient???_frame??.nii.gz')):\n",
    "\n",
    "                        # Save information about patient\n",
    "                        with open(os.path.join(folder_path, 'Info.cfg')) as f:\n",
    "                            lines = f.readlines()\n",
    "\n",
    "                        ED = int(lines[0].strip()[-2:])\n",
    "                        ES = int(lines[1].strip()[-2:])\n",
    "\n",
    "                        # Split file name\n",
    "                        file_base = file.split('.nii.gz')[0]\n",
    "                        file_mask = file_base + '_gt.nii.gz'\n",
    "\n",
    "                        # Load data from .nii.gz files\n",
    "                        img_nii = nib.load(file)\n",
    "                        img_dat = img_nii.get_fdata()\n",
    "\n",
    "                        mask_nii = nib.load(file_mask)\n",
    "                        mask_dat = mask_nii.get_fdata().astype(int)\n",
    "                        \n",
    "                        # if(np.logical_or('patient038' in file , 'patient057' in file)):\n",
    "                        #     print(np.unique(mask_dat))\n",
    "\n",
    "\n",
    "                        img = img_nii.get_fdata()\n",
    "                        mask = mask_nii.get_fdata()\n",
    "\n",
    "                        pixel_size = img_nii.header.get_zooms()\n",
    "\n",
    "                        # Make vector to make all images have the same resolution\n",
    "                        scale_vector = [pixel_size[0] / target_resolution[0], pixel_size[1] / target_resolution[1]] \n",
    "\n",
    "                        for zz in tqdm.tqdm(range(img.shape[2])):\n",
    "\n",
    "                            # Normalize, rescale and crop the image and  mask\n",
    "\n",
    "                            slice_img = np.squeeze(img[:, :, zz])\n",
    "                            slice_img = normalize_img(np.squeeze(img[:, :, zz]))\n",
    "                            #slice_img = np.squeeze(img[:, :, zz])\n",
    "                            img_rescaled = transform.rescale(slice_img,\n",
    "                                                             scale_vector,\n",
    "                                                             order=1,\n",
    "                                                             preserve_range=True,\n",
    "                                                             mode='constant')\n",
    "\n",
    "                            slice_mask = np.squeeze(mask[:, :, zz])\n",
    "                            # slice_mask = normalize_img(np.squeeze(mask[:, :, zz]))\n",
    "                            mask_rescaled = np.around(transform.rescale(slice_mask,\n",
    "                                                              scale_vector,\n",
    "                                                              order=0,\n",
    "                                                              preserve_range=True,\n",
    "                                                              mode='constant'))\n",
    "                            #print(np.unique(mask_rescaled))\n",
    "                            img_cropped = crop_pad_resize(img_rescaled, nx, ny)\n",
    "                            mask_cropped = crop_pad_resize(mask_rescaled, nx, ny)\n",
    "                                                       \n",
    "                            \n",
    "                            # if(np.logical_or('patient038' in file , 'patient057' in file)):\n",
    "                            #     #print(np.unique(mask_rescaled))\n",
    "                            #     # print(file_base)\n",
    "                            #     corrupted_files.append(file_base)\n",
    "                            # else:\n",
    "                            \n",
    "                            # Find largest connected component:\n",
    "                            img_cropped,mask_cropped=get_ROI(img_cropped,mask_cropped,target_resolution)\n",
    "                                \n",
    "                            if denoise:\n",
    "                                img_cropped = denoise_tv_chambolle(img_cropped, eps=1e-6, weight=alphaTV, max_num_iter=1000)\n",
    "                            \n",
    "                            # if not (np.logical_or('patient038' in file , 'patient057' in file)):\n",
    "                            # Save images in PNG format\n",
    "                            if 'frame{:02}'.format(ED) in file:\n",
    "                                img_loc = os.path.join(foldername+'/'+train_test, file[len_inp:-7]+'_slice{:01}_ED'.format(zz)+'.png')\n",
    "                                img_fin = Image.fromarray(np.uint8(255*img_cropped),mode=\"L\")\n",
    "                                img_fin.save(img_loc, format='PNG')\n",
    "\n",
    "                                mask_loc = os.path.join(foldername+'/'+train_test, file[len_inp:-7]+'_slice{:01}_ED_gt'.format(zz)+'.png')\n",
    "                                mask_fin = Image.fromarray(np.uint8(mask_cropped), mode=\"L\")\n",
    "                                #print('uniquecrop'+ str(np.uint8(np.unique(mask_cropped))))\n",
    "                                mask_fin.save(mask_loc, format='PNG')\n",
    "                            else:\n",
    "                                img_loc = os.path.join(foldername+'/'+train_test, file[len_inp:-7]+'_slice{:01}_ES'.format(zz)+'.png')\n",
    "                                img_fin = Image.fromarray(np.uint8(255*img_cropped), mode=\"L\")\n",
    "                                img_fin.save(img_loc, format='PNG')\n",
    "\n",
    "                                #print(np.shape(mask_cropped))\n",
    "                                mask_loc = os.path.join(foldername+'/'+train_test, file[len_inp:-7]+'_slice{:01}_ES_gt'.format(zz)+'.png')\n",
    "                                mask_fin = Image.fromarray(np.uint8(mask_cropped),mode=\"L\")\n",
    "                                mask_fin.save(mask_loc, format='PNG')\n",
    "                else:\n",
    "                    print('Folder for '+folder_path[len_inp:]+' is not empty. No files were written to this folder.')\n",
    "        \n",
    "    # print(corrupted_files)\n",
    "    print(\"Preprocessing Finished\")\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ebbca9d3-6919-4562-b86c-dd2e32944bac",
   "metadata": {},
   "source": [
    "#### Preprocess the data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "4c785ee5-f3a6-4c21-a1bc-ab3776dca597",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 10/10 [00:00<00:00, 54.58it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 54.11it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 54.13it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 55.40it/s]\n",
      "100%|██████████| 11/11 [00:00<00:00, 49.23it/s]\n",
      "100%|██████████| 11/11 [00:00<00:00, 49.48it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 51.52it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 42.94it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 54.31it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 56.26it/s]\n",
      "100%|██████████| 6/6 [00:00<00:00, 56.72it/s]\n",
      "100%|██████████| 6/6 [00:00<00:00, 55.91it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 60.23it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 60.88it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 55.58it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 55.95it/s]\n",
      "100%|██████████| 7/7 [00:00<00:00, 56.28it/s]\n",
      "100%|██████████| 7/7 [00:00<00:00, 55.71it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 40.27it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 62.17it/s]\n",
      "100%|██████████| 18/18 [00:00<00:00, 58.17it/s]\n",
      "100%|██████████| 18/18 [00:00<00:00, 57.66it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 58.26it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 57.26it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 41.95it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 59.79it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 29.37it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 57.92it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 58.71it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 58.68it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 52.88it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 51.23it/s]\n",
      "100%|██████████| 7/7 [00:00<00:00, 49.16it/s]\n",
      "100%|██████████| 7/7 [00:00<00:00, 48.99it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 55.78it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 59.18it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 58.14it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 60.30it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 24.23it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 54.63it/s]\n",
      "100%|██████████| 13/13 [00:00<00:00, 53.79it/s]\n",
      "100%|██████████| 13/13 [00:00<00:00, 50.46it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 61.07it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 59.10it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 54.85it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 56.71it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 51.83it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 59.98it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 58.97it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 56.88it/s]\n",
      "100%|██████████| 12/12 [00:00<00:00, 62.98it/s]\n",
      "100%|██████████| 12/12 [00:00<00:00, 59.34it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 56.71it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 56.98it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 54.68it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 55.46it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 63.21it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 58.10it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 37.15it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 43.60it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 54.87it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 55.54it/s]\n",
      "100%|██████████| 7/7 [00:00<00:00, 53.78it/s]\n",
      "100%|██████████| 7/7 [00:00<00:00, 36.20it/s]\n",
      "100%|██████████| 7/7 [00:00<00:00, 19.37it/s]\n",
      "100%|██████████| 7/7 [00:00<00:00, 58.94it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 57.80it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 51.59it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 52.21it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 51.80it/s]\n",
      "100%|██████████| 15/15 [00:00<00:00, 32.47it/s]\n",
      "100%|██████████| 15/15 [00:00<00:00, 60.02it/s]\n",
      "100%|██████████| 16/16 [00:00<00:00, 60.37it/s]\n",
      "100%|██████████| 16/16 [00:00<00:00, 54.65it/s]\n",
      "100%|██████████| 7/7 [00:00<00:00, 58.09it/s]\n",
      "100%|██████████| 7/7 [00:00<00:00, 58.93it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 61.66it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 58.17it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 30.67it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 49.14it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 59.98it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 59.04it/s]\n",
      "100%|██████████| 7/7 [00:00<00:00, 53.39it/s]\n",
      "100%|██████████| 7/7 [00:00<00:00, 51.40it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 58.13it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 57.72it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 52.41it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 53.38it/s]\n",
      "100%|██████████| 16/16 [00:00<00:00, 55.89it/s]\n",
      "100%|██████████| 16/16 [00:00<00:00, 34.27it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 54.18it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 51.97it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 55.03it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 55.34it/s]\n",
      "100%|██████████| 14/14 [00:00<00:00, 60.08it/s]\n",
      "100%|██████████| 14/14 [00:00<00:00, 58.70it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 55.83it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 55.91it/s]\n",
      "100%|██████████| 7/7 [00:00<00:00, 31.95it/s]\n",
      "100%|██████████| 7/7 [00:00<00:00, 54.92it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 52.85it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 51.89it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 52.41it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 50.83it/s]\n",
      "100%|██████████| 14/14 [00:00<00:00, 57.35it/s]\n",
      "100%|██████████| 14/14 [00:00<00:00, 57.68it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 46.47it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 56.77it/s]\n",
      "100%|██████████| 6/6 [00:00<00:00, 56.08it/s]\n",
      "100%|██████████| 6/6 [00:00<00:00, 41.52it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 60.33it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 60.45it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 56.76it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 48.29it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 56.99it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 57.31it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 52.96it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 49.59it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 59.79it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 53.90it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 48.27it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 52.72it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 58.01it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 49.96it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 55.23it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 53.88it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 58.10it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 57.63it/s]\n",
      "100%|██████████| 17/17 [00:00<00:00, 55.79it/s]\n",
      "100%|██████████| 17/17 [00:00<00:00, 45.31it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 58.16it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 58.76it/s]\n",
      "100%|██████████| 16/16 [00:00<00:00, 60.42it/s]\n",
      "100%|██████████| 16/16 [00:00<00:00, 55.79it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 51.38it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 55.38it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 51.63it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 41.40it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 59.60it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 51.81it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 57.78it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 50.77it/s]\n",
      "100%|██████████| 12/12 [00:00<00:00, 38.20it/s]\n",
      "100%|██████████| 12/12 [00:00<00:00, 50.92it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 48.79it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 52.44it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 50.22it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 53.50it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 47.93it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 49.04it/s]\n",
      "100%|██████████| 7/7 [00:00<00:00, 54.80it/s]\n",
      "100%|██████████| 7/7 [00:00<00:00, 55.98it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 54.90it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 47.00it/s]\n",
      "100%|██████████| 6/6 [00:00<00:00, 53.05it/s]\n",
      "100%|██████████| 6/6 [00:00<00:00, 53.11it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 54.76it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 52.88it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 56.59it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 54.29it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 53.16it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 55.41it/s]\n",
      "100%|██████████| 15/15 [00:00<00:00, 54.61it/s]\n",
      "100%|██████████| 15/15 [00:00<00:00, 44.42it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 53.75it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 51.73it/s]\n",
      "100%|██████████| 6/6 [00:00<00:00, 47.74it/s]\n",
      "100%|██████████| 6/6 [00:00<00:00, 49.89it/s]\n",
      "100%|██████████| 7/7 [00:00<00:00, 53.21it/s]\n",
      "100%|██████████| 7/7 [00:00<00:00, 54.65it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 53.53it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 56.32it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 55.23it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 54.10it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 55.21it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 27.83it/s]\n",
      "100%|██████████| 7/7 [00:00<00:00, 54.39it/s]\n",
      "100%|██████████| 7/7 [00:00<00:00, 56.65it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 46.52it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 46.26it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 42.93it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 49.49it/s]\n",
      "100%|██████████| 11/11 [00:00<00:00, 52.68it/s]\n",
      "100%|██████████| 11/11 [00:00<00:00, 55.70it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 54.44it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 33.95it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 51.56it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 53.47it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 39.09it/s]\n",
      "100%|██████████| 8/8 [00:01<00:00,  7.67it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 49.36it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 52.16it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 45.82it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 46.66it/s]\n",
      "100%|██████████| 6/6 [00:00<00:00, 51.43it/s]\n",
      "100%|██████████| 6/6 [00:00<00:00, 52.14it/s]\n",
      "100%|██████████| 11/11 [00:00<00:00, 56.68it/s]\n",
      "100%|██████████| 11/11 [00:00<00:00, 56.59it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 49.50it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 33.42it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 53.79it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 48.66it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 61.46it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 55.79it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 58.29it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 52.64it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 54.57it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 53.92it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 47.81it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 47.05it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 29.85it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 49.45it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 55.62it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 56.45it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 59.08it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 61.57it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 51.58it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 59.17it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 59.73it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 60.70it/s]\n",
      "100%|██████████| 16/16 [00:00<00:00, 46.25it/s]\n",
      "100%|██████████| 16/16 [00:00<00:00, 56.61it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 55.02it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 56.28it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 62.32it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 52.64it/s]\n",
      "100%|██████████| 17/17 [00:00<00:00, 60.09it/s]\n",
      "100%|██████████| 17/17 [00:00<00:00, 55.99it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 27.35it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 57.63it/s]\n",
      "100%|██████████| 19/19 [00:00<00:00, 56.36it/s]\n",
      "100%|██████████| 19/19 [00:00<00:00, 60.46it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 61.53it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 58.08it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 55.90it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 56.81it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 59.11it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 58.30it/s]\n",
      "100%|██████████| 6/6 [00:00<00:00, 62.65it/s]\n",
      "100%|██████████| 6/6 [00:00<00:00, 58.91it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 57.27it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 55.50it/s]\n",
      "100%|██████████| 20/20 [00:00<00:00, 56.31it/s]\n",
      "100%|██████████| 20/20 [00:00<00:00, 44.26it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 55.61it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 52.66it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 13.70it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 59.24it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 59.05it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 57.66it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 58.44it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 57.30it/s]\n",
      "100%|██████████| 11/11 [00:00<00:00, 57.85it/s]\n",
      "100%|██████████| 11/11 [00:00<00:00, 59.45it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 53.10it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 53.72it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 31.32it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 57.03it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 49.00it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 55.50it/s]\n",
      "100%|██████████| 21/21 [00:00<00:00, 55.90it/s]\n",
      "100%|██████████| 21/21 [00:00<00:00, 51.45it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 51.93it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 53.08it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 43.51it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 49.65it/s]\n",
      "100%|██████████| 15/15 [00:00<00:00, 47.76it/s]\n",
      "100%|██████████| 15/15 [00:00<00:00, 50.29it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 59.87it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 26.34it/s]\n",
      "100%|██████████| 16/16 [00:00<00:00, 57.20it/s]\n",
      "100%|██████████| 16/16 [00:00<00:00, 59.11it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 33.60it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 54.59it/s]\n",
      "100%|██████████| 11/11 [00:00<00:00, 57.47it/s]\n",
      "100%|██████████| 11/11 [00:00<00:00, 55.10it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 58.58it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 56.69it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 56.25it/s]\n",
      "100%|██████████| 8/8 [00:00<00:00, 56.34it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 58.41it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 59.43it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 21.40it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 56.03it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 55.26it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 49.96it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 51.55it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 52.68it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 56.46it/s]\n",
      "100%|██████████| 9/9 [00:00<00:00, 58.63it/s]\n",
      "100%|██████████| 17/17 [00:00<00:00, 58.66it/s]\n",
      "100%|██████████| 17/17 [00:00<00:00, 45.09it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 54.26it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 54.46it/s]\n",
      "100%|██████████| 17/17 [00:00<00:00, 53.84it/s]\n",
      "100%|██████████| 17/17 [00:00<00:00, 60.66it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 53.09it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 55.33it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 54.32it/s]\n",
      "100%|██████████| 10/10 [00:00<00:00, 42.56it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Preprocessing Finished\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "target_resolution = (1.36719, 1.36719)\n",
    "target_size = (256, 256)\n",
    "data_path = '../ACDC/database/'\n",
    "\n",
    "#preprocess(data_path, target_resolution, target_size)\n",
    "preprocess(data_path, target_resolution, target_size, denoise=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8a240090-a4ba-43d3-838d-78462f1f2e9a",
   "metadata": {},
   "source": [
    "#### Denoise the data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "9cc831d0-c58b-492d-bac7-dbdf0d0f6965",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "done\n"
     ]
    }
   ],
   "source": [
    "# preprocess(data_path, target_resolution, target_size, denoise=True, alphaTV=0.2)\n",
    "print('done')\n",
    "# alpha = 0.1 geeft wel prima maar nog veel details, 0.3 is misschien net hoog."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "3b06a56f-d604-4810-a7ea-f87a2a8a9528",
   "metadata": {},
   "outputs": [],
   "source": [
    "# ik heb alpha =0.1 gekozen, target size veranderd naar 256*256 --> werkt beter icm onze kernal afmetingen en aantal lagen van netwerk, voor gecropte images gebruik ik een target size van 128*128\n",
    "\n",
    "# Voor ROI alpha verlaagd naar 0.05 ---> toch naar 0.1"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}